Geostatistiske

Geostatistics is a branch of statistics that deals with spatial or spatiotemporal data, aiming to quantify spatial variation and to predict values at locations where measurements are not available. In Norwegian usage, the term geostatistiske refers to methods, models, and results associated with geostatistics.

The discipline emerged in the 1960s through the work of Georges Matheron and colleagues, who introduced the concept of a regionalized variable and the variogram as a tool to describe spatial dependence. Core ideas include spatial autocorrelation, stationarity assumptions (second-order stationarity or intrinsic hypothesis), and the use of models for the spatial covariance or semivariogram.

Interpolation and prediction are typically carried out with kriging, a family of linear unbiased estimators that combine a known data point with the spatial correlation structure. Variants include ordinary, universal, and indicator kriging, as well as co-kriging and regression-kriging. Conditional simulation can be used to quantify uncertainty. Geostatistics also covers non-Gaussian and non-stationary settings through trend modeling and inclusion of covariates.

Data are usually point measurements, but block data and gridded surfaces are common in practice. Analyses rely on robust experimental designs, variogram estimation, cross-validation, and uncertainty assessment.

Geostatistical methods are applied in mining and mineral exploration, hydrogeology, environmental science, agriculture, petroleum, and epidemiology. Popular software tools include GSLIB, SGeMS, and R packages such as gstat and geoR, as well as GIS platforms with geostatistical extensions.